Skip to content
View shriyansnaik's full-sized avatar

Highlights

  • Pro

Block or report shriyansnaik

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
shriyansnaik/README.md

Shriyans Naik

AI Engineer · LLM and agentic systems

Portfolio · LinkedIn · Email


I build LLM systems end to end: QLoRA fine-tuning and vLLM serving, retrieval, and multi-agent platforms that run in production. Currently working on an agentic compliance platform at Big Language Solutions. Previously Swiggy (founding AI engineer on CREW, a chat-based concierge platform), National Stock Exchange, and Jio Platforms.

Mumbai, India.

Minions

PyPI Python versions

minion-ai is an agentic framework I wrote because every other one made me bolt observability on separately, and the good answer was always a paid hosted service. Here, tracing of every run, turn and tool call, plus per-model cost tracking, is one flag.

import minions

minions.init(tracing=True, project="demo")
agent = minions.Minion(model="openai/gpt-4o", tools=[get_weather])
print(agent("Should I take a jacket in Oslo today?"))

A tool is a plain function; its signature and docstring are the schema. Traces go to a local SQLite file you own, with a dashboard you run yourself (minion serve) and spend roll-ups across a 145-model price table. Sub-agents, parallel tool calls, and a strict JSON envelope per turn that keeps agents portable across OpenAI, Anthropic, Gemini, Azure and vLLM through LiteLLM.

It runs the entire agent layer of the compliance platform at work, all 25 agents.

Docs · Quickstart · Cookbook

Currently

  • Working into the agents, evals and RL-environments side of open source, starting with PrimeIntellect-ai/verifiers
  • Pushing Minions toward a 1.0

Selected work

minion-ai Agentic framework with tracing, cost tracking and a self-hosted dashboard. On PyPI.
multimodal_rag RAG over PDFs that indexes figures as well as text, by summarising every extracted image and embedding it alongside the prose.
linkedin-feed-parser Chrome extension that turns LinkedIn job posts into structured cards via LLM extraction, and drafts the outreach email.
shribook Event finance tracking for singing groups. React, Firebase, role-based access, live here.
google-translate-api Selenium driver for Google Translate behind a FastAPI wrapper. Chunking, 190+ languages, no official API.

Toolbelt

Languages Python · Go · SQL · JavaScript

LLM Agentic systems · MCP · RAG · Structured outputs · Fine-tuning (QLoRA, PEFT) · Evals

Frameworks LangGraph · DSPy · LiteLLM · vLLM · Ray · PyTorch · Transformers · FastAPI · Pydantic

Data Qdrant · Milvus · Elasticsearch · PostgreSQL · MongoDB · DynamoDB · Redis · Snowflake

Infra AWS · Azure · Docker · GitHub Actions · Langfuse · Grafana · Prometheus

Pinned Loading

  1. minion-ai minion-ai Public

    A simple agentic framework - observability, evals, memory all in one

    Python 4

  2. linkedin-feed-parser linkedin-feed-parser Public

    Chrome extension that parses LinkedIn job posts into clean cards with LLM extraction, flip-card UI, and Gmail email drafting

    JavaScript